which is downloadable from
http://sanskritlibrary.org/downloads.html has a large array of transcoders one of which transcodes to Romanization using precomposed Unicode characters that include diacritics. The problem with searching that Harry Spier mentions is just one of a number of reasons why Malcolm Hyman and I designed the Sanskrit Library phonetic encoding for all our linguistic programming, including both the encoding of texts and searching, and use Unicode only for display, and data input if desired (though for the latter purpose SLP and most other meta-encodings are preferable). Our book
Linguistic Issues in Encoding Sanskrit available at
http://sanskritlibrary.org/publications.html discusses the issues comprehensively.